Historical GitHub Uptime Charts
82 points by hibachrach
82 points by hibachrach
Sure, they may have made the uptime worse, but remember what we got in exchange -- when it's up, the UI is slower and buggier.
They also used all our code to train Copilot. And we didn’t get bothered by questions like whether we’re OK with that.
Even if you go to breakdown and remove "issues", "actions" (this seems the most volatile), and "pages", it still looks pretty bad.
I wouldn't remove issues TBH (it's a core forge function after all). Also, excluding copilot & codespaces makes it greener than it actually is.
I recall a time where seeing the GitHub unicorn daily was a thing. This was before Microsoft.
Any idea why this isn’t represented here? Could it be that this unicorn period was even before 2016?
I think giving transient/random unicorns to a low fraction of requests is not considered outage from the statuspage point of view, neither before nor under MS.
Yes, GitHub went thru an extremely unstable period during its early growth, but I think in this case the problem is that the data source just fails to indicate when data is missing:
https://www.githubstatus.com/uptime?page=3000
Here it claims there was 100% uptime is 1996.
I don't remember this much except on large issues/PRs and even then their support was pretty top notch right up until the acquisition.
I'm skeptical about this data and the suggested interpretation. The trend seems unrealistically pronounced to me. It can't be that uptime was at a perfect 100% for years and then dropped immediately and permanently from the moment of the Microsoft acquisition. This data may be ignoring some axes that are actually necessary for an accurate interpretation.
For a start, you can't factor in Actions uptime into overall uptime. That product feature just didn't exist at all before Microsoft acquired GitHub, so it's not fair to say that Microsoft made it worse.
Yeah a possible confounding factor is "they changed how the reporting is done". I could believe it though, given that they also migrated to Azure at some point. Just rebuilding the amount of infrastructure necessary for that is probably a pretty big deal.
Also note the Y axis: it's 99.5% to 100%.
They did not migrate to Azure at some point, they are currently migrating to Azure.
Ooh, thanks for the correction. Might still be related? Though since it looks like they announced that October 2025 and the problems started roughly October 2019, probably not.
I'd like to see this somehow normalized to total Github size. Their stats only go back to 2020, but between then and now the number of hosted respositories tripled.